Portfolio

Portfolio

Claude Code

Swift SDK for MetaWear — GitHub

Hardware SDKs for the MetaMotion wireless sensor platform. Full BLE transport layer, protocol implementation, all sensor modules (accelerometer, gyroscope, magnetometer, sensor fusion, barometer, GPS, LED, haptic, GPIO, timer, events, macros, serial passthrough, data processor chains, firmware OTA).

896 unit tests. Hardware integration tests on real iOS devices. iOS app with modern design (iOS 17+ Liquid Glass), Demo Mode (no hardware required), 3D orientation visualization. Structured as 4 SwiftPM products with SwiftData persistence and CSV export.

Kotlin SDK for MetaWear — GitHub

Same feature set as Swift. Kotlin 2.1.20, minSdk 26, KMP-ready architecture separating protocol (pure JVM) from transport and persistence (Android).

1,275+ JVM tests. 43 BLE + 13 persistence instrumented tests, hardware-verified on Pixel 7 with MetaMotion S. Jetpack Compose app with feature parity to iOS. Includes DemoBleTransport — a protocol-level emulator for testing without hardware.

Both shipping, in active use at scale.

Claude MCP

MCP Server: Career Timeline & Posts — mcp.laura.kassovic.com

Static site MCP server exposing career data via standard MCP tools. Schemas for get_career_timeline (all roles), get_role (by company), list_posts (all posts), list_posts_for_era (posts by era), get_about (bio). Runs on Jekyll; data sourced from _data/journey.json and _posts/. Tool discovery via /mcp.json manifest. Enables Claude or any MCP client to programmatically query structured career information.

Claude Skills

Beacon Support Skill — 1,624 Discord threads + 402 email threads. Safety policies baked in.

MbientLab Support Skill — Multi-language API support (Swift, Kotlin, C++, Python). 520+ real support messages indexed and searchable.

Gradle Skills — github.com/gradle/gradle-skills

Built and shipped with team. Two production skills: gradle-best-practices and gradle-wrapper-upgrade. Each skill is a self-contained instruction set with:

  • Clear step-by-step instructions for agents
  • Input/output schemas validated per step
  • Error handling and rollback logic
  • Testing against 7 LLM vendors
  • Metrics for success (accuracy, speed, safety)

gradle-best-practices audits a build against Gradle’s best practices (naming, organization, plugin usage, performance anti-patterns), reports violations, proposes fixes. Tested on real builds. Claude Sonnet 5 improved from 30 to 50 of 51 checks with skill enabled.

gradle-wrapper-upgrade handles wrapper upgrades (download, pin checksum, verify build, rollback on failure). Tested on matrix of Gradle versions and build configurations.

Design: Each skill trades off prescriptiveness (more instructions = less agent confusion) against flexibility (less instructions = agent can adapt). Both skills document why each instruction exists and what agents struggle with without it.

Claude Agents

Daily To-Do Aggregator — Local Python server

Agent that runs every morning, aggregates to-dos from multiple sources (Slack, GitHub, emails), and creates a consolidated to-do list. Built with Claude API. Demonstrates multi-source integration, scheduled execution, and agentic decision-making on filtering and prioritization.

PR Documentation Agent — Triggered on every PR to gradle/gradle

Built with Claude API and a custom Gradle Skill. On every new PR to the main Gradle repository, the agent examines code changes, queries existing documentation, and automatically creates new docs, new upgrade guide entries, new release notes, or updates existing versions as needed. Catches what needs documenting without human intervention. Solves field friction: code ships faster than docs. Demonstrates content-freshness automation, agentic multi-step reasoning, and feedback loop (code change → intelligent doc update).

Gradle Contribution Agents — github.com/gradle/gradle/.agents

Co-authored and maintain 5 developer guidance agents in the Gradle repository: API Changes, Building, Code Review, Testing, and Tooling. Each agent is a structured instruction set that helps contributors navigate complex processes:

  • API Changes — Guidance on public vs. internal API, nullability annotations (@NullMarked, @Nullable, @since versioning), binary compatibility checks with sanityCheck.
  • Building — Build commands, project structure (platforms, subprojects), language levels, avoiding full-build pitfalls (./gradlew build), best practices for targeting and compilation.
  • Code Review — Reviews follow the Code Review guide—scope detection, correctness bugs, API violations, edge cases, security; excludes style and formatting. Output: structured findings with severity levels.
  • Testing — Spock test framework, unit vs. integration tests (forkingIntegTest, configCacheIntegTest), helpers, patterns for testing Gradle builds; linked to GitHub issues.
  • Tooling — Git commit message rules (7-rule format), IDE integration (IntelliJ diagnostics), relative paths for code navigation.

Demonstrates agentic guidance systems at scale: teaching agents (and humans) how to contribute correctly, reducing friction on large codebases, and embedding domain expertise into repeatable instruction sets.

GTM & Partnerships

Gradle Security Subscription Launch — blog.gradle.org

Led full product launch for paid security subscription. Coordinated with product on feature messaging (SLA-based fixes, extended maintenance, embargoed advance notice). Built enablement materials for sales. Worked directly with GTM teams and sales to pitch to enterprise customers (organizations with compliance obligations, running legacy Gradle versions). Direct partnership on customer calls and deal progression:

Strategic Work

Gradle Is Going Agentic — blog.gradle.org, August 2026

Co-authored with Tom Tresansky. Positions Gradle for agents as first-class users. Agents increasingly run builds, read failures, propose fixes. Gradle’s approach: official skills (teach agents the right patterns), benchmarks (measure whether skills work), friction removal (clearer errors, less token waste).

AI Policy — github.com/gradle/gradle

Co-authored. Guidelines for AI-assisted contributions to Gradle. Requires contributors to understand and justify their submissions. Disclosure of significant AI involvement. Human review and judgment as final gate on all contributions. Enforcement against low-effort AI-generated content and non-engagement in review.

Feature Evaluation & Product Feedback — GitHub Issues

Documentation process requires stress-testing features like a real user. Write release notes, which means identifying edge cases, confusing behavior, and gaps. File formal feedback via GitHub issues with a:bug and a:feature labels. Because team is small, also provide direct Slack feedback to engineers. Demonstrates understanding of what developers struggle with before features ship to the field, and closes the feedback loop from user experience back to product.

Training, Support & Enablement

DPE University — dpeuniversity.gradle.com

Gradle’s hands-on training platform. Built course automation pipeline: Claude generates presentation slides + hands-on project + instructor script. moss-tts-8b synthesizes narration with voice cloning (your voice, not generic TTS). Assembled in Camtasia/DaVinci Resolve. Best seen on YouTube

Course creation: 1 week → 1–2 days (5–7x faster). Covers Gradle fundamentals, build optimization, enterprise patterns.

Gradle Community Support - Daily support on Gradle community forum and Slack.

Answer technical questions, help developers debug build issues, provide architecture guidance, and identify patterns that feed back into documentation improvements. Direct signal from the field — understanding what developers struggle with and where the gaps are.

Gradle Documentation Impact - Rebuilt Gradle documentation architecture and content organization.

Measured effectiveness through CSAT (customer satisfaction) scores. Post-rebuild improvement in developer satisfaction directly correlates to reduction in support friction and improved onboarding velocity. Demonstrates commitment to measuring enablement effectiveness beyond “content exists”—showing actual impact on user experience and satisfaction.

Public Speaking & Advocacy

GitHub Universe 2025 — San Francisco, October 28, 2025

Co-presented with Tom Tresansky on “Avoiding the next supply chain disaster using GitHub and Gradle.” Covered the Dependency Submission Action, dependency scanning, and how to secure supply chains at scale. Blog post.

Gradle Build Cache Webinar — develocity.ai, July 2025

Co-hosted with Akshay Dayal. Covered local and remote build caching in Gradle: how reuse works, cache hit analysis, Develocity Build Cache for CI and shared machines, analytics tools, and certification through DPE University. Free online event for developers.

API:World 2024 — Santa Clara, CA, November 6, 2024

Co-presented with Baruch Sadogursky on “OpenAPI Extensibility: The Good, The Bad and The YAMLy” in the Developer Productivity Engineering track (3:00–3:25 pm PST). Session page.

MIOTherapy Launch — Business Wire, January 28, 2019

Quoted as co-founder and CEO on MbientLab’s physical therapy product launch. Coverage picked up by ODT Magazine, Medical Design and Outsourcing, and HIT Consultant.

Sensors Expo & Conference — Speaker Schedule, 2018

Spoke on MetaWear wearable sensor platform capabilities, wireless protocols, real-world IoT applications.

Medical Sensors Design Conference — Speaker Schedule, 2018

Presented on miniaturized wearable sensors, medical-grade BLE design, data accuracy in motion sensing for health applications.

IoT World 2018 — Santa Clara, 2018

MbientLab participated in Project Kairos startup program with Startup City showcase. ReadWrite coverage; also noted on Parsers.vc.

AEC Business Interview: “Sensors for Smarter Construction” — Podcast, November 2016

Discussed three areas where wearables improve construction sites: worker safety monitoring (real-time health signals), insurance (accident data collection), workflow efficiency (hands-free data entry). MbientLab manufacturing model and sensor reliability at scale.

Press & Interviews

MetaWear Coin Kickstarter — Wareable

Featured as CEO discussing MetaWear’s mission: “what Arduino did for schools, we want to do for wearables.”

MetaWear Launch — Make Magazine

Interviewed on the MetaWear board and its Kickstarter campaign.

MetaWear Kickstarter Coverage — Neoteo (Italian)

Featured as the engineer behind the MetaWear board.

Arm Innovator Series — Hackster.io Video

Interview on MetaWear platform: size and washability improvements, Bluetooth 5 capabilities for long-range wireless sensors, power efficiency gains, real-world deployment in health and fitness.

Reusable Demo Environments

Gradle Ephemeral CI Docker Images — Part 1, Part 2, Part 3

Three-part series on optimizing Gradle builds in ephemeral CI with ready-to-use Docker images. Provided Dockerfiles for primed Gradle environments with pre-populated caches, dependency caches, and build cache strategies. One-command setup (docker build/docker pull) for developers and CI systems to run consistent, optimized Gradle builds without manual configuration. Demonstrates building reusable, reproducible environments at scale.